Claude Vs. ChatGPT - Bubble knowledge

thatdevguy · 2024-11-27T21:26:22.827Z

Hey all,

Small tip I wanted to share - Over the past year I’ve worked heavily with both Claude and ChatGPT as part of my workflow. Typically this includes any custom code I need to write, along with brainstorming design aspects and workflow setup. I’m certain many of you do the same.

However, a lot of times I have questions that are specific to Bubble. With Bubble undergoing major UI and feature upgrades over the years I notice that A LOT of responses to bubble specific questions are typically wrong or just fully hallucinated by the models. Yes, even Claude which is renowned for it’s ability to handle advanced tasks & coding/engineering tasks.

That was until recently. About 2-3 months back it seems that ChatGPT has VERY ACCURATE knowledge of the current Bubble UI and platform in general. So much so that I’ve almost stopped reading through the forums completely as the solution is almost always correct with ChatGPT.

I really hope to see the same with Claude eventually maybe once Anthropic includes web browsing and does a little fine tuning.

TLDR: ChatGPT > Claude for Bubble platform knowledge

cha_james · 2024-11-27T22:02:03.232Z

I’ve noticed this too! Running ideas or questions past either is super helpful now, I commonly just dump API doc links in chatgpt and use it to clarify call setup

thatdevguy · 2024-11-27T22:19:51.784Z

Nice! When prompting, if you provide really detailed context in your prompt about whatever problem you’re trying to solve (like 2 paragraphs worth, no joke) and ask for a step by step solution, the results are incredible. I think a lot of that has to do with ChatGPT’s web browsing functionality so it’s able to search official up to date Bubble docs along with crowdsourced solutions such as those answered/discussed in the forums. However, context is huge when prompting. You get much better results when starting with one long prompt vs. multiple questions in a single discussion.